Thinking Machines Lab unveils full duplex AI interaction models

Thinking Machines Lab, the artificial intelligence startup founded by former OpenAI Chief Technology Officer Mira Murati, announced on Monday the introduction of what it calls “interaction models.” These models are designed to process user input and generate responses simultaneously. This approach represents a shift away from the turn-based structure of conversational AI, where a system must wait for a user to finish speaking before it begins processing a response. Instead, the startup is developing a model that enables “full duplex” interaction—a technical term for a system that can process incoming information and output responses at the same time. In practice, this means the AI can listen and talk concurrently, allowing for real-time, simultaneous communication where a user could interrupt the machine mid-sentence, much like a natural human conversation.

According to Thinking Machines Lab, its model, named TML-Interaction-Small, achieves a response time of roughly 0.40 seconds. This speed is intended to mimic the natural cadence of human conversation. By processing input and generating output concurrently, the model aims to bypass the turn-based latency that characterizes current conversational systems. This latency is a standard feature of existing models developed by competitors in the field, including OpenAI and Google, which rely on a sequential “listen-then-respond” architecture. In those sequential systems, the interaction feels more like a text-based exchange, where each party must wait for the other to finish before speaking.

The technology is still in its development phase and is not yet available for general public use. Thinking Machines Lab has stated that a limited research preview is coming in the next few months, with a wider release planned for later this year. This phased rollout will allow the company to test the model’s real-world performance and determine whether its simultaneous processing capabilities can consistently deliver a natural, phone-call-like conversational experience outside of a controlled research environment. The company is positioning the technology as a research preview rather than a fully realized consumer product, meaning that its practical utility will only be verified once users can interact with it directly.

Why it matters

The artificial intelligence industry is attempting to transition from turn-based, text-chain-style interactions to full duplex systems that process input and output simultaneously. If successful, this architectural shift could make voice interactions with AI feel as natural and fluid as a standard human phone call.